stratified sampling
Stratified sampling is splitting a group into smaller layers before you survey or sample it, so every layer, more than the loud ones, gets fairly counted.
In plain English
Here is a mistake a lot of small business owners make without realising it. A shop owner in Ikeja wants to know if customers would pay ₦2,000 more for a premium version of her product, so she asks the twenty customers who walked in that Saturday afternoon. She never stops to ask who those twenty people actually are. Maybe they are mostly students because Saturday afternoon is when students come out, and mostly students will say no to the higher price, so she scraps the idea, when actually her weekday market women customers would have said yes without blinking.
That is the trap. You think you sampled your market, but you actually just sampled whoever was standing nearest to you.
Stratified sampling fixes this by making you slice your population into layers first, based on something that actually matters to your question, like age, income, location, or how often they buy. Then you deliberately pick a fair number of people from each layer, more than from the layer that happens to be loudest or closest.
So our shop owner, if she had done this properly, would split her customers into students, working professionals, and market women, then ask a fair number from each group, more than whoever wandered in during her lunch break.
This matters hugely for entrepreneurs testing an idea, because a small business cannot afford to build a product for the wrong crowd based on a lopsided glimpse of the market. A UK example would be a startup testing a new banking app only on university WhatsApp groups, then wondering why retirees hate the interface, because retirees were never in the sample to begin with.
Even someone in London sending money home can spot this logic. If you asked only your friends in Lagos whether a transfer app's fees feel fair, you would miss what your aunty in Kano or your cousin in Enugu actually experiences, and their reality matters just as much.
A startup wants feedback from 500 customers spread across three states. Instead of grabbing the first 500 who reply on Instagram, it pulls 200 from Lagos, 150 from Abuja, and 150 from Port Harcourt, matching each region's real share of its customer base, so the ₦0 cost of a badly designed survey does not turn into a ₦2,000,000 mistake in product decisions later.
Questions people ask
How is stratified sampling different from just picking people randomly?
Plain random sampling treats everyone as one big pool, so by pure luck you might end up asking mostly one type of person. Stratified sampling deliberately breaks the pool into meaningful groups first, then samples fairly within each group, so no group gets accidentally ignored.
Do small business owners really need this, or is it just for big researchers?
Small businesses need it just as much, arguably more, because they cannot afford a wrong read on their market. A trader guessing prices based on ten random customers can lose real money, while spending an extra hour grouping customers properly before asking questions costs nothing but attention.
What is a simple way to start using stratified sampling in my own business?
Identify one or two things that genuinely divide your customers, like location, age range, or spending level, then make sure your survey or feedback list includes a fair spread across those groups before you draw any conclusions from it.
Last reviewed